Interoperability through semantic labeling with context Andrzej Bialecki

Interoperability through semantic labeling with context Andrzej Bialecki <abial@webgiro. com> ECIMF meeting, Paris 2002. 06. 10

Semantics: the key to interoperability n n n Proper understanding of the meaning of business, service and protocol data is crucial Most e-commerce standards use incompatible semantics for their key concepts Business context affects the concepts semantics Most e-commerce standards can be represented as ontologies (shared conceptualizations of given knowledge domains) There is an acute need for context-aware semantic translation (mapping) techniques

Semantic translation techniques n Direct translation n n Unification and adoption n Requires N*(N-1) translations “from scratch” – a popular Q’n’D approach Requires a synthesis of N ontologies – non-trivial! Requires replacement of N ontologies with 1 common – impossible! Multiple ontologies local ontology n Requires just N translations (plus N*(N-1) refinements… ) Requires a comprehensive yet abstract shared ontology – nontrivial local ontology Single ontology Global ontology Multiple ontologies + labels Shared ontology Labeling n local ontology

Role of context n Local context n n Relationships to other concepts in the same ontology External context n n Relationships to concepts outside the ontology Examples: n n a “purchase” term is commonly understood by people in general a “cargo unload at port of call” term is commonly understood by people involved in shipping of goods

Semantic labeling with context n n n Attach labels taken from shared ontology to your concepts Find corresponding labels in the foreign ontology Apply more steps to refine the relationships: n n n Local context Automated, formal reasoning and inference External context – semantic enrichment Heuristics (best practice and rule of thumb ) Define the translation rules in a formal way

Semantic labeling in action – importance of context e. g. eb. XML cc. DRIV Context drivers’ values Shared ontology Process. procurement. q uoting Party. identification. details. p erson Rosetta. Net PIP 3 A 1: Request Quote PIP 3 A 2: Request Price and Availability Quote. Request. from Role Quote. Request. to. Ro le e. g. UBL, BSR, or eb. XMLCC EDIFACT Business transactions and documents Business Information Entities REQOTE / QUOTES NA D

Interoperability gains n Better understood business semantics of key concepts n n This influences system integration implementations Significant reduction of mapping efforts n Instead of M*N*(N-1) mappings between N standards in M contexts, you can start with M*N mappings to a common universal ontology

Work to be done n n Further research into methods and techniques Conversion of existing standards into precise machine-readable models n n EDIFACT – XML version of DIRDEF needed! Excel and Word docs are pain! n n UML models are sometimes too much! n n n Incompatible XMI versions, the format itself is very complex, vague UML semantics DTD, XSD or RDF models often seem to fit the bill Construction of upper-level shared ontology n n Non-portable, often poorly structured and difficult to process automatically MULECO draft, BSR, UBL, eb. XML-CC … ? Tools supporting the methodologies?

Application to ECIMF n n n The ECIMF Semantic Translation Tool uses semantic labeling with context The tool is extensible and modular, so that experience can be gained with this approach and if it proves inappropriate, it can be changed Initial testing using a subset of eb. XMLCC shows interesting results

More information n ECIMF project n n CEN/ISSS WS-EC n n http: //www. cenorm. be/isss Web. Giro AB n n http: //www. ecimf. org http: //www. webgiro. com Author: n Andrzej Bialecki <abial@webgiro. com>